I feel like these breaches started to accelerate after Anthropic wrote a full post about how easy it is to abliterate glm 5.3 to make a frontier-level cyber model.

Now, to be clear, abliteration was no state secret, but that blog post was pure accelerationism.

▲

tptacek 5 hours ago | parent | next [-]

That was just 30 days ago. This has been happening for months. Part of the "Black October" narrative is just vulnerability research; Calif didn't abliterate 5.3 to find PHP2Shell.

▲

postalcoder 4 hours ago | parent | next [-]

Yeah IIRC calif research on php2shell preceded all of that but I still stand by the thought that Anthropic has been the most accelerationist AI company despite its lip service to the notion of safety.

▲

tptacek 4 hours ago | parent [-]

I don't have a take about Anthropic generally. The blog post was quite good. The blog post did not cause October's security debacles.

▲

postalcoder 4 hours ago | parent [-]

It was a good blog post yes, but I think you’re underplaying the bannister effect of Anthropic coming out and endorsing glm 5.3’s cyberattack prowess.

	▲	tptacek 4 hours ago \| parent [-]
		The idea that attackers were waiting for Anthropic to tip them off on this is kind of risible. It's been a huge topic amongst practitioners for the better part of a year, and practitioners are usually a step or two behind dedicated criminal or state-sponsored attackers.

▲

_notreallyme_ 5 hours ago | parent | prev [-]

They are using mythos-5.5, which is still better than an abliterated GLM-5.3.

	▲	tptacek 4 hours ago \| parent [-]
		Right, that's kind of my point. Also: it's a good blog post (the Anthropic one), but it observes abliterated 5.3 models were already public. It's not like people who break into computers didn't know that already. They're not waiting for Anthropic to tell them about open weight models; they already live in the world of open weight models.

▲

nater5000 3 hours ago | parent | prev | next [-]

Duh, the idea that those capable of actually utilizing these models for these purposes weren't aware that they could use these models for these purposes up until Anthropic released a blog post about it is silly. Anybody who is even a bit tech literate has been well aware of this inevitability for basically as long as open-weight models have been a thing.

The timing of all of this is much easier to pin to the creation and proliferation of capable of open weight models. Anthropic calling this out is not the cause for it; it's just an obvious correlation.

▲

paxy 4 hours ago | parent | prev | next [-]

lol no, abliteration communities have existed and abliterated models routinely published long before Anthropic wrote that post last month.

What’s changing now is that models are getting more powerful.

▲

sandeepkd 4 hours ago | parent | prev [-]

You are running a machine with tons of vendor provided agents/scripts to do god knows what on a cloud provider which itself runs tons of agents/scripts to manage that infrastructure and then you expect to be secure.

And no, before the labs there was a degree of fear for everyone to hack into some one else system. With the labs changing the narrative, anyone can just blame the agents for the breach, whoever did it would only get fame and no blame